Biosocial determinants of healthy ageing in Spain

(1)

(2)

(3)

PhD Thesis

Biosocial determinants of healthy ageing in Spain

Author:

Santiago Rodr´ıguez L´ opez

Supervisor:

Pilar Montero L´ opez

Departamento de Biolog´ıa Facultad de Ciencias

Universidad Aut´ onoma de Madrid ESPA ˜ NA

2014

(4)

(5)

Abstract

This dissertation explores biological (physical) and social characteristics associated with health later in life, integrating gender/sex considerations and a life course perspective. We use data from 50-years-old and older Spanish and other European adults from two different sources: the Active Ageing Longitudinal Study “Estudio Longitudinal de Envejecimiento Activo” (ELEA) in Spain, and the Survey of Health, Ageing and Retirement in Europe (SHARE). This study includes cross-sectional and longitudinal (both prospective and retrospective) designs and embraces a broad set of variables such as demographic characteristics, socioeconomic and health indicators, early health and socioeconomic conditions, etc.

Results are presented in five original papers. Cross-sectionally, we found that, compared to men, part of the overall poorer health among women was surely determined by a gender effect and may have had an early origin, probably related to traditional gender roles established early in life. Additionally, we found marked socioeconomic gradients in health and mobility indicators like frailty and balance performance. Health behaviours like physical activity and obesity seemed to play a similar and small role in explaining the link between socioeconomic status and frailty and balance in older adults. Longitudinally, in a prospective study of the predictors of disability for two years, we found that a decline in function was associated with an increased number of chronic diseases and symptoms of depression among Spanish men, whereas among women it was associated with decreased cognitive performance. Finally, in a retrospective study with a life course approach, we found a direct association between childhood and adult health among older Europeans, whereas the impact of the socioeconomic status in childhood was more indirect and operated through the own socioeconomic status in adulthood. This suggests that in order to improve adult health, efforts can be made in ameliorating child health. Moreover, poor childhood health was a stronger predictor of adult health -having more nega-

(6)

tive effect- in Northern compared to other European countries. This finding may be useful for planning interventions based on country-specific evidence, and contributes to the understanding of the mechanisms underneath the health dynamics over the life course.

The results of this study add to the evidence of the importance of including a multidisciplinary and life course perspective when evaluating health and well- being in later life. They might also contribute to enhance health and reduce health inequalities by suggesting effective interventions meant to improve the quality of life of older adults.

Key words: Healthy ageing; adult health; biosocial determinants; gender;

life course; social inequalities; ELEA; SHARE; Spain; Europe.

4

(7)

Abstract (Spanish)

Esta tesis explora caracter´ısticas biológicas y sociales asociadas a la salud en la edad adulta, integrando consideraciones de género-biolog´ıa y una perspectiva de ciclo vital. Se utilizan datos de individuos españoles y europeos mayores de 50 años de edad provenientes de dos fuentes: el Estudio Longitudinal de Envejecimiento Activo (ELEA) en España y la Encuesta de Salud, Envejeci- miento y Jubilación de Europa (Survey of Health, Ageing and Retirement in Europe; SHARE). Este estudio incluye diseños transversales y longitudinales (prospectivos y retrospectivos), abarcando un amplio grupo de variables tales como aspectos demográficos, indicadores socioeconómicos y de salud, salud y condiciones socioeconómicas tempranas, etc.

Los resultados se estructuran en cinco art´ıculos originales. Los estudios transversales mostraron que parte de la peor salud general de las mujeres en relación a los hombres estar´ıa determinada por un efecto de género y podr´ıa tener un origen temprano, relacionado probablemente a los roles tradicionales de género establecidos a temprana edad y condicionado mayormente por el acceso a la educación. Además, encontramos marcados gradientes socioeconómicos en distintos indicadores de salud y mobilidad como fragilidad y equilibrio. Aunque asociados a peores indicadores de salud, malos comportamientos de salud como la inactividad f´ısica o la obesidad parecen jugar un papel similar con una con- tribución relativamente menor en explicar las desigualdades en fragilidad y equilibrio en los adultos mayores españoles. Por otra parte, estudiando los predic- tores de discapacidad a lo largo de dos años en el estudio longitudinal prospec- tivo, observamos que una pérdida en funcionalidad f´ısica estaba asociada a un aumento de enfermedades y s´ıntomas depresivos entre los hombres españoles, mientras que en las mujeres se asociaba a una disminución del funcionamiento cognitivo. Finalmente, en el estudio retrospectivo con un enfoque de ciclo vital, encontramos una asociación directa entre la salud temprana y adulta en adultos

(8)

europeos, mientras que el impacto del nivel socioeconómico en la niñez era más indirecto, operando mediante el propio nivel socioeconómico en la edad adulta.

Esto sugiere que para mejorar la salud en etapas posteriores del ciclo vital, parte de los esfuerzos deber´ıan destinarse a mejorar la salud temprana. Además, una pobre salud en la niñez resultó un mayor predictor de salud (teniendo un efecto más negativo) en los pa´ıses del norte comparados con otros pa´ıses europeos.

Este resultado puede ser ´util para desarrollar intervenciones basadas en eviden- cias particulares y contribuye al conocimiento de los mecanismos subyacentes a las din´amicas de salud a lo largo del ciclo vital.

Los resultados de este estudio a˜naden a la evidencia existente acerca de la importancia de incluir una perspectiva multidisciplinar y de ciclo vital a la hora de estudiar la salud y bienestar en la edad adulta. Tambi´en puede contribuir a mejorar la salud y disminuir las diferencias en la misma, sugiriendo intervenciones efectivas dirigidas a mejorar la calidad de vida de los adultos mayores.

Palabras clave: Envejecimiento saludable; salud adulta; determinantes bioso- ciales; g´enero; ciclo vital; desigualdades sociales; ELEA; SHARE; Espa˜na; Eu- ropa.

6

(9)

Acknowledgements

This study was carried out at the Department of Biology of the Universidad Aut´onoma de Madrid (UAM). Part of it has also been developed during two research visits at the Department of Public Health, Section of Social Medicine, University on Copenhagen, Denmark, and at Max Planck Institute for Demo- graphic Research, Research Group on Life Course Dynamics and Demographic Change, Rostock, Germany. I had the privilege of working with many highly skilled and wonderful persons who have given their valuable contribution to the study.

First of all, I would like to express my deepest appreciation to my supervisor, Pilar Montero L´opez PhD, for guiding me during these years. Her expertise, constructive advice, and encouragement have had invaluable influence on my work and future. Thank you!

I want to thank all the coauthors of the original papers for your collaboration:

Mauricio Avendano PhD, Charlotte Nilsson MD PhD, Rikke Lund MD PhD, Professor Emeritus Roc´ıo Fernández-Ballesteros PhD, Mar´ıa Dolores Zamarrón PhD, Aldana González Montoro PhD, Margarita Carmenate PhD. In addition, my very special thanks are dedicated to the supervisors during the research visits, Professor Mikko Myrskylä PhD, and to the memory of that lovely women we have recently lost, Professor Kirsten Avlund DMSc.

I would also like to express my sincere gratitude to the official reviewers of this thesis, Professor Sonia E. Colantonio PhD and Professor Charles Susanne for your comments and thorough evaluation of the thesis.

I am grateful for the financial support that I have received for completing my thesis. My doctoral studies and research were financially supported by a position at the UAM and by personal grants funded by the UAM and Ministry of Education of Spain. In addition, I wish to thank all the workers and participants in the projects carried out during this period for giving their essential share for this study.

(10)

In addition, my very special thanks are dedicated to my colleagues at the Unit of Anthropology of the UAM for providing inspiring and supportive working environment.

My deepest gratitude I want to express to my family, specially to parents, Estela and Roberto, and to my aunt Eli. I cannot thank enough for your encouragement and support throughout my life and during the process of this study. My dearest thanks go to Al for her support and patience during these busy years. Love you all.

Madrid, February 2014.

Santiago Rodr´ıguez L´opez

8

(11)

List of original publications

The study is based on the following original publications:

I Montero López P, Fernández-Ballesteros R, Marrodán Zamarrón MD, Rodr´ıguez López S. (2011). Anthropometric, body composition and health determinants of active ageing: A gender approach. Journal of Biosocial Science, 43:597-610.

II Rodr´ıguez López S, Nilsson C, Lund R, Montero López P, Fernández- Ballesteros R, Avlund K. (2012). Social inequality in dynamic balance performance in an early old age Spanish population: The role of health and lifestyle associated factors. Archives of Gerontology and Geriatrics, 54:e139-e145

III Rodr´ıguez L´opez S, Montero L´opez P, Carmenate Moreno M. Educational inequalities and frailty in Spain: What is the role of obesity?. Accepted for publication: The Journal of Frailty & Aging (January 2014)

IV Rodr´ıguez L´opez S, Montero L´opez P, Carmenate Moreno M, Avendano M.

(2013). Functional decline over 2 years in older Spanish adults: Evidence from the Survey of Health, Ageing and Retirement in Europe. Geriatrics

& Gerontology International, doi:10.1111/ggi.12115.

V Rodr´ıguez López S, Myrskylä M, González Montoro AM, Montero López P. The long-term health implications of poor childhood health: Evidence of regional variation from the Survey of Health, Ageing and Retirement in Europe. Submitted for publication: Population Studies (February 2014)

(12)

(13)

Abbreviations

95% CI 95% Confidence Intervals

AA Active Ageing

AIC Akaike Information Criterion ADL Activities of Daily Living ANOVA Analysis of Variance ATPIII Adult Treatment Panel III BIC Bayesian Information Criterion BMI Body Mass Index (kg/m²)

CEMFI Centro de Estudios Monetarios y Financieros (Spain) ELEA Estudio Longitudinal de Envejecimiento Activo ELSA English Longitudinal Study of Ageing

EURO-D European Depression Scale

FD Functional Decline

HA Healthy Ageing

HRS Health and Retirement Study

IADL Instrumental Activities of Daily Living

IECM Instituto de Estad´ıstica de la Comunidad de Madrid (Spain) INE Instituto Nacional de Estad´ıstica (Spain)

ISCED International Standard Classification of Education

LE Life Expectancy

MMSE Mini Mental State Examination

OR Odds Ratio

PUMA Programa Universitario de Mayores

SD Standard Deviation

SES Socioeconomic Status

SHARE Survey of Health, Ageing and Retirement in Europe SHARELIFE Wave 3 of SHARE

(14)

Continue...

SRHS Self-Reported Health Status W(1-4) Waves(1-4) of SHARE WHO World Health Organization

12

(15)

Introduction

1.1 Population ageing and the health paradox in Spain

Population ageing is a major challenge for all countries in Europe. The age structure of the population has changed over recent decades and there has been an unprecedented increase in the number of older people. According to the European Union Prospective “Confronting demographic changes, solidarity between generations”(2008), the population over 75 years will increase from 26.4 million in 2000 to 45.3 in 2030 due to ageing population, who starred the baby boom of the 70’s and the enhance in life expectancy (LE) beyond the seventy- five years. In Spain in 2006, more than 7 million people were over 65 years and 1,300,000 over 80. In our country, between 1980 and 2005, male LE at birth has increased from 72.5 to 77.0 years and female from 78.6 to 83.5 years [1], while the LE free of disability is around 7 years for the total population. Hence, an important question is whether the above mentioned increases in LE will increases LE free of disability (the “increase”scenario) or if on the contrary, new policies intending to promote healthy ageing and prevent disability will have a positive impact in reducing it (hypothesis of “reduction of morbidity”) [2, 3].

Spain is one of the European countries where high LE and low mortality rates coexist with a very high level of functional disability. This health issue is relevant within the described context and evaluating the determinants of this disability may contribute to explore answers to this paradox.

(18)

1.2 Studying healthy ageing: Active Ageing and its bio/psycho/social frame

There is an individual variability in the ageing process that causes important physiological differences among individuals with the same chronological age [2, 4]. Owing to this discrepancy between chronological and biological or physiological age, it is difficult to define what normal ageing is. The widely accepted notion of successful ageing was established by Rowe & Kahn in 1997 [5].

They proposed a multidimensional concept called Active Ageing (AA), defined by several biological, psychological and social factors, which is based on the lack of illnesses or disability, good physical and cognitive functioning, and active social participation. They classified the many ways of ageing into three categories:

common, pathological and successful ageing, although other terms are similarly used to describe successful ageing, such as healthy ageing [6], apt ageing [2, 7], active ageing [8, 9] and productive ageing [10].

In our study we use the concept of AA to represent the terms referring to successful ageing. AA is a scientific concept that requires a multidisciplinary research and most authors agree on the fact that the determinants of AA must be studied from a bio-psycho-social perspective. The collaboration between professionals from different areas is important to achieve an effective promotion of AA which takes into account the biological, psychological and sociocultural determinants involved in this complex process.

AA is associated with different factors in men and women. Some studies have shown that, among women, AA is negatively related to physical and physiological variables associated with poor health, whereas in men it seems that social factors such as educational level and occupation may have more influence. Other studies have shown great differences in the prevalence of AA among older adults, depending on the domains considered to define it [11, 12].

More recently, McLaughlin et al. [13] estimated the prevalence of healthy aged North-Americans in 12%. In Europe, a comparative cross-sectional study following Rowe and Kahn’s criteria have shown prevalences of AA ranging from 21.1% in Denmark to 1.6% in Poland [14]. In the case of Spain this prevalence is 3.1%, one of the lowest among all the European countries. However, this approach has also been criticised. It was suggested that the comparison of the prevalences of active agers across studies is of limited use given the wide variety of definitions and measurement approaches [12].

These cross-national differences are partially due to the environmental, economic, cultural and social conditions in a particular historical context that affect

16

(19)

the ageing process. Historical disparities among these characteristics between men and women are more accentuated in older ages. Thus, the current objective of health initiatives to increase LE free of disability in the next years can only be achieved by fully incorporating a gender perspective in population studies and by considering different measures for men and women.

1.3 Gender/sex integrated aspects in health and welfare research

The main factors behind the demographic changes mentioned above are related to economic and social development: declining fertility rates and increasing LE are related to improvements in living conditions of the population and medical care that have affected women and men differently [8]. The control of mortality associated to childbirth and puerperium has been one of the factors that has increased LE in women, but there are other factors responsible for inequalities that would be accountable for the observed differences in the life quality of older men and women. Although women have about 7 years longer LE than men, their LE free of disability is lower [15]. Therefore, the gender perspective is essential when analysing health of the older European population.

In general, both objective and perceived health get worse throughout life and the gender-based approach is of special interest in this sense, since women have poorer health than men. The World Health Organization (WHO) states that the gender position during the ageing process is a factor that conditions adult health [8]. Particularly in Spain, four out of ten women over 65 classified their health as poor or very poor, whereas this occurred in three out of ten men [1]. Although women live longer than men, only 65% of them reported living in good health, compared to the 70% declared by men. The different perception related to health might be due to gender inequality in employment and economic, personal and leisure autonomy. Mortality and morbidity rates and the trends of specific diseases among older people also differ from men and women. Most of these differences might be rooted in the development conditions in early and later life, since throughout their life course, men and women are exposed to different risk factors that influence their health in later years [16].

Most of the participants in the present study were born between 1935 and 1950. At that time, Spain was one of the poorest countries in Europe. There have been changes in the political system (1930-36), the occurrence of the Civil War (1936-39) and the postwar period (1939-50), coinciding with the Second

(20)

World War, which led to the isolation of Spain, food shortages and other negative events, with the corresponding increase of mortality and morbidity. Older people from this generation might be considered “survivors” and their bio-physical and psychological characteristics need to be considered within this historical context. This highlights the necessity of having a deep knowledge of the interaction between gender, health and ageing.

The biological ageing process is characterised by a progressive decrease in functional capacity in all tissues and organs of the body, and by the decreased ability to respond and adjust to environmental changes [4]. At the individual level, ageing is a process that has no precise beginning, it occurs throughout the life and depends on genetic, biological and psychological factors [2, 5, 17]. The physiological changes that occur throughout the ageing process involve a remod- elling in the size, shape and body composition [18]. These variations in body size are mainly due to a loss of height -due to compression of the intervertebral discs, to the loss of bone mass and the loss of plantar arch curvature [19].

Body composition is affected by a decrease in metabolically active lean body mass, due to loss of muscle mass (sarcopenia) [20] and of cells from different tissues and organs, as well as from the skeleton demineralisation [21]. The reduction of lean mass in skeletal muscle results in a loss of strength and greater fatigue which can lead to the abandonment of simple activities and to the increased risk of falls. These changes in body composition are associated with functional decline (FD), disability and morbidity [22]. Not only would these factors have a negative influence on life quality by increasing the degree of dependence, but also they would enhance the risk of mortality and morbidity [23].

The changes in body composition associated with ageing have different consequences in men and women. Differences in body composition between men and women exist since puberty. There is a higher prevalence of fat mass in women than in men -25% vs. 15%- and greater muscle and bone mass in men.

Women have higher calcium requirements associated with pregnancy and lacta- tion and the differences in calcium metabolism mediated by oestrogens produce a rapid loss of bone mass in the years following menopause. The loss of bone mass in women is higher than in men and may lead to osteoporosis problems and fractures more frequently than in men. Among women, it is very common to experience obesity and sarcopenia simultaneously. Moreover, body fat distribution -different between men and women- is also a determinant of cardiovascular risk. All these factors are influenced by the environmental conditions throughout the life course: primarily nutrition and energy expenditure associated with daily physical activity heavily depend on the gender position, especially among

18

(21)

population groups whose living conditions were unfavourable in the early years of their life.

The use of the Body Mass Index (BMI) -considered as a good indicator of underweight, overweight and obesity- is controversial among older individuals. This is mainly due to difficulties in obtaining accurate height values and also because BMI does not take into account neither the total amount of fat nor its distribution [24]. However, other measures such as span -the distance from the breastbone to the tip of the middle finger, the arm circumference and subcutaneous fat skinfolds can be used to obtain more accurate information. In addition, many studies have shown an association between higher values of BMI with lower mortality risk in men and women over 70 years, but abdominal obesity is particularly associated with an increased risk of cardiovascular disease.

In this study we use BMI from both measured and self-reports of weight and height, despite the large debate on whether this last is a reliable indicator of the nutritional status. In this line, many studies have found a general overestima- tion of height and an underestimation of weight, resulting in an underestimation of BMI [25, 26].

1.4 The life course perspective when studying health in adulthood

Inequalities between older men and women continue to increase with age. These inequalities can only be explained from a life course perspective and as a result of the interaction between biological and social factors over the life course.

Particularly, it was suggested that the study of adult health cannot be fully accomplished without considering the exposure to the different environmental conditions throughout the life course [27].

In the last years there has been renew interest for explaining causality pat- terns between both childhood socioeconomic (SE) and health conditions and their consequences in adult life -in health, SE achievement, etc. [28–32]. Con- sequently, many studies have focused on the association between life course SE conditions with different health outcomes in adulthood [33–38]. One of the main arguments of those studies is that the social environment during the growth period is strongly associated with the health conditions accumulated over the life course and therefore, affect health in adulthood [39].

Although adult SE conditions remain the most commonly addressed aspects of health disparities [32], large evidence suggests that conditions early in life

(22)

have long-term effects on health at older ages [40]. Thus, childhood health, nutritional status, socioeconomic status (SES), place of residence and other household characteristics also contribute to disparities in adult health [27, 35].

There is evidence on two major and sometimes conflicting models to explain how early life environment influences health in later life [41]. Among the mechanisms through which childhood environment influences adult health, there is the direct effect of childhood into adult health [42]. This can be described as a latency model, which emphasises the strong independent effects on health status late in life, of discrete events that tend to occur early in life [41]. The associations between birthweight, placenta size and weight gain in the first year of life with cardiovascular disease in the fifth decade have been described [43]. It is hypothesised that, in the case of the latter associations, future adult disease is

“programmed” during foetal life and infancy.

On the other hand, there are other indirect mechanisms operating through attained adult characteristics (e.g., SES and lifestyle factors) [44]. This can be described as a pathways model, which emphasises the role of early environment on subsequent life trajectories, which in turn influence adult health. In other words the pathways model focuses on the cumulative effect of life events along developmental trajectories, and it thereby implicates conditions of life throughout the life course in adult disease causation [41, 45]. Figure 1.1 shows the potential mechanisms linking childhood and adult health described above that are addressed in this study.

The importance of distinguishing between these two mechanisms relies on contributing to the study of health trajectories over the life course, leading to suggest efficient interventions to improve adult health.

The existing evidence about childhood conditions as good predictors of adult health, together with the differential exposure to health risk factors along the life course in men and women [46], may suggest that gender differences in later life health have an early origin. Thus, in order to improve later life health, it is important to detect the early determinants of adult health and to identify how health trajectories are represented in both men and women from different environments.

20

(23)

Figure 1.1: Potential mechanisms linking childhood and adult health.

1.5 New contributions to the study of the de- terminants of health and well-being in older adults

There is a persistent interest in researching the determinants of health in older adults, being specially important in the Spanish population due to the context mentioned above. The present study is structured based on five original papers which intend to summarise the relevance of undertaking the study of adult health from a multidisciplinary and life course perspective, by considering biological and social integrated aspects.

In Paper I (4.1) we evaluate, from a gender approach, how different physical and social characteristics are associated with AA among Spanish adults. It is known that men and women are exposed to different risk factors along the life course that affect their health in later years [47] and that this differential exposure is often due to different gender roles established early in life [16]. In this study we use a combined indicator of health like AA, which has been proposed

(24)

to embrace the bio-psycho-social dimension of healthy ageing [8]. Moreover, the United Nations Research Agenda on Ageing for the 21^st Century stated the need for research into the determinants of this type of positive ageing as a priority [48]. Both Papers I (4.1) and II (4.2) are based on cross-sectional data from the Active Ageing Longitudinal Study “Estudio Longitudinal de En- vejecimiento Activo”(ELEA) -developed by the Psychobiology Group of the Universidad Aut´onoma de Madrid (See Table 2.1).

Papers II (4.2) and III (4.3) provide somehow similar cross-sectional evidences on the role of different health behaviours in explaining health inequalities in adulthood. It has been previously described how social inequalities in health are distributed in the Spanish population [49]. However, the role that health-related behaviours might play in such association is less known. Paper II (4.2) focuses on educational and income inequalities in an objective measure of mobility like dynamic balance, and how objective obesity and physical activity are associated. Balance performance is considered a good indicator of adult health and one of the most important determinants to protect independence later in life [50], while poor balance performance is an important cause of loss of independent mobility [51]. This paper was elaborated during a research visit to the Department of Public Health, Section of Social Medicine, University of Copenhagen, Denmark, and supervised by Prof. Kirsten Avlund^†.

Papers III (4.3) and IV (4.4) use data of the Spanish sample from the Survey of Health, Ageing and Retirement in Europe (SHARE). First, Paper III (4.3) investigates educational differences in the prevalence of frailty phenotypes -an indicator of health status in old age and a good predictor of disability in adulthood [52]- and the specific role of obesity in such association. Previous research examined the association between SES and frailty [53–55], while others investigating the relationship between obesity and frailty have been comparatively minor [56, 57]. Few studies have examined, however, how obesity is related to frailty within different educational backgrounds. The importance of determining the role of obesity in the development of frailty has been recently suggested [58], since probably the most common phenotype of frailty in the near future will be characterised by the concurrent and interacting presence of obesity [59]. In a recent study, Macklai et al. [60] found that a SHARE’s frailty phenotype [52]

is significantly associated with large health outcomes, independent of baseline morbidity and disability in community-dwelling European men and women aged 60 and older. The robustness of these results and others [61] validate the use of this phenotype in SHARE for future research on frailty in Europe.

Paper IV (4.4) introduces the longitudinal perspective in our study using SHARE’s data. We include a longitudinal analysis of disability and FD in the

22

(25)

Spanish population, by evaluating social, educational, health and behavioural predictors of physical FD. Earlier research have examined associations between health and disability measures [62], but few studies have examined how broader social, educational and behavioural determinants are related to FD in Spain.

Here we provide a comprehensive examination of the predictors of changes in FD in older Spanish adults based on longitudinal data. Our study is innovative by examining how both baseline and 2-year changes in predictors relate to changes in FD, and by assessing a wide array of potential predictors across multiple domains.

Finally, Paper V (4.5) is based on SHARE’s international panel data, including ten European countries. We evaluate how poor health in childhood is associated with individuals’ later life health within different European regions, assessing whether both the exposure to different SE situations over the life course and health risk behaviours in adulthood mediated such association, over a 6.6-year window. A growing body of evidence in the last years suggests that conditions early in life have long-term effects on health at older ages [63–65].

However, most of the studies on the multiple determinants of health across the life course do not include cross-national/regional comparisons and, to our knowledge, our study provides the first regional-specific evidence of historical differences in the childhood-adult health association. This paper was elaborated during a research visit to the Max Planck Institute for Demographic Research, Research Group Life Course Dynamics and Demographic Change, Rostock, Ger- many, and supervised by Prof. Mikko Myrskyl¨a.

(26)

(27)

Objectives

The determinants of health in older adults should be studied from a multidisciplinary and life course perspective. Our research hypothesis is that the gender/biology interaction remains all along the life course and distinctively conditions later life health in men and women. This study aims to evaluate how biological and social factors are associated with health later in life, incorporating a gender approach into the study of health in adulthood, allowing for the integration of health and environment research considerations throughout the life course. Within this context, the specific objectives of this study are:

1. Evaluate how biological (physical) and socioeconomic (educational attain- ment, income, profession, etc.) determinants of health are associated with active ageing in Spanish men and women.

2. Assess SE disparities in adult health indicators such as frailty and balance in Spain, estimating the contribution of modifiable health behaviours on such inequalities.

3. Estimate SE, health and behavioural predictors of physical functional decline among Spanish adults.

4. Evaluate, from a life course perspective, the importance of childhood conditions in determining later life health among older European adults.

(28)

(29)

Material and Methods

2.1 Study design and participants

Data from two different sources were used in this study: the Active Ageing Longitudinal Study “Estudio Longitudinal de Envejecimiento Activo”(ELEA) and the Survey of Health, Ageing and Retirement in Europe (SHARE). Both samples are complementary, and provide different perspectives to this study.

The use of two data sets independently, enhances the scope of the analyses by allowing the use of complementary variables and by performing cross-sectional, longitudinal and retrospective analysis. Table 2.1 resumes the data used in every paper.

2.1.1 Active Ageing Longitudinal Study: the ELEA project

The ELEA was developed by the Psychobiology Group of the Universidad Aut´onoma de Madrid in 2006. It has a longitudinal design, although at the moment of the present study only the baseline sample was available. Baseline sample is formed by 456 -65 to 80 years- men and women from rural and urban areas of the Comunidad Aut´onoma of Madrid (Madrid and Toledo), Spain. It also includes a sub sample from the PUMA (n=24). Participants were inter- viewed and measured in homes and in older adults community institutions.

One strength of the ELEA is that it includes many anthropometric indicators, physical performance tests and other objectively measured health variables.

On the other hand, the cross-sectional outline somehow restricts the scope of our study. Overall, it represents a good source of data which is used in Pa- pers I (4.1) and II (4.2) of this study. See those papers for more details on the methodology and data collection.

(30)

2.1.2 The Survey of Health, Ageing and Retirement in Europe (SHARE)

SHARE is a multidisciplinary and cross-national panel database of micro data on health, SES and social and family networks of more than 85,000 individuals from 19 European countries. It includes population-based data on economic, social, and health conditions for Spanish and other European individuals. Based on probability samples in all participating countries, SHARE represents the non-institutionalised population aged 50 and older. The survey has been designed and harmonised following the Health and Retirement Study (HRS) of North America and the English Longitudinal Study of Ageing (ELSA).

The Centre of Monetary and Financial Studies (Centro de Estudios Monetar- ios y Financieros (CEMFI)) is the SHARE’s responsible institution in Spain (www.share.cemfi.es).

At the moment of the present study SHARE has released four waves. Base- line sample (W1) was recruited in 2004/05, while further waves such as the second (W2), third (W3, SHARELIFE), and fourth (W4) were released in 2006/07, 2008/09 and 2011/12, respectively. With the inclusion of W2 and further waves, SHARE went into its longitudinal dimension, allowing for different approaches.

SHARELIFE includes retrospective data on early SE and health indicators, to elucidate how early life experiences and events throughout life influence the circumstances of older people. SHARELIFE included a Life History Calendar, designed to help respondents in remembering past events more accurately. The use of the life history calendar technique has been shown to improve the accuracy of the retrospective information given by respondents [66]. See www.share- project.org for more specific information on the methodology.

In our study we use baseline (W1) (Paper III (4.3)) and longitudinal (W1- W2) (Paper IV (4.4)) data for Spain. Moreover, we include longitudinal (W1- W3-W4) data for ten European countries (Paper V (4.5)), including Denmark, Sweden, Germany, Switzerland, The Netherlands, Austria, France, Belgium, Italy and Spain (see Table 2.1 for more details on this).

2.2 Outcomes

Most of the outcomes considered in this study are a combination of several variables (e.g. Active Ageing, Dynamic Balance, Frailty phenotype, and Func- tional Decline). The decision on focusing on these types of indicators relies on the differences between the samples considered in each paper. Compared to

28

(31)

when evaluating a specific health outcome, the use of combined outcomes gives a somehow broader perspective of the health status in each particular analysis.

Contrarily, in many cases these outcomes do not provide for specific information and restricts the contrast and comparison of the findings by limiting the chances of its replication. Despite this limitation, we consider that the use of combined indicators provides a broader notion of the general health status of the studied population.

2.3 Statistical analysis

A somehow similar methodology was followed in every paper. First, some descriptive statistics and bi-variant associative analysis -χ², t-test, ANOVA, etc.- were performed, followed by different predictive models. We used linear, negative binomial, logistic, or multinomial logistic regression analysis depending on the studied outcome. Non-parametric analysis were used when appropriated.

Since odd ratios (ORs) values are sometimes difficult to interpret when effects are protective, in most studies we present them as non-protective [67]

(ORs>1.00). So, the lowest risk category in every case is selected as the joint/reference category, resulting in a positive difference for higher risk categories.

In addition, SHARE includes different sets and types of weights, which can be used depending on the concrete research question [68]. Particularly, in this study we have used calibrated longitudinal weights in Paper IV (4.4). These weights are only defined for the longitudinal sample and compensate for problems of attrition between two or more waves. Thus, they are calibrated to match, e.g., the target population of W1 that survives in W2, so they also account for mortality, which is a phenomenon affecting both the sample and the population [69].

Statistical analysis were performed using Stata Statistical Software [70], the Statistical Package for Social Sciences [71] and R [72].

(32)

Table2.1:Studydesigns,populationsandoutcomesStudyDatasetDesignParticipantsAge(mean±SD)Primaryoutcomes

IELEACross-sectionalSpain54-75(66.5±5.4)ActiveAgeing456IIELEACross-sectionalSpain54-75(66.4±5.3)DynamicBalance448IIISHARECross-sectionalSpain50-103(66.8±10.6)FrailtyPhenotype2,319IVSHARELongitudinalSpain65-103(74.3±6.4)FunctionalDecline2-yearfollow-up699VSHARELongitudinalTencountries65-103(63.4±8.8)Self-reportedhealth6.6-yearfollow-up7,118ChronicconditionsRetrospectiveGripstrength

30

(33)

Results and Discussion

3.1 Summary of the results

This section provides a brief summary of the main findings from the original papers. The first results describe the factors associated to AA in Spanish men and women (Paper I (4.1)). Then, we summarise the role of different health behaviours on educational and income inequalities in health indicators such as dynamic balance and frailty (Papers II (4.2) and III (4.3), respectively). Then, we address the predictors of physical FD in Spain in a longitudinal analysis (Paper IV (4.4)). Finally, we present longitudinal evidences on the childhood- adult health association and the regional differences observed across Europe (Paper V (4.5)).

3.1.1 Gender differences in active ageing in Spain (Paper I)

In Paper I (4.1) we analyse, from a gender perspective, how physical, health and SE factors are associated with AA in Spain. Our sample is formed by 456 community-dwelling Spanish adults (169 men and 287 women), mean age 66.5 years old (range 54-75 years).

To describe the factors associated with AA, we created a dichotomous vari- able “active ageing” (AA) (yes/no), based on the main dimensions of AA reported by Depp and Jeste [12]: cognitive and disability/illness/physical functioning, subjective health, satisfaction with life and productive activity performed.

We found AA to be associated with anthropometric variables, physical health indicators and socio-demographic characteristics. Men have better general health than women. The prevalence of active agers is higher in men than in women

(34)

(38.4% vs. 21.9%; p<0.001), and compared to women, men have lower prevalence of obesity (29.0% vs. 37.6%; p<0.01), significantly lower metabolic risk (42.9% vs. 60.6%; p<0.001) and better self-perceived health (19.5% of men reported very good health vs. 11.8% of women; p<0.005).

Table 3.2 shows the main results based on multiple logistic regression analysis. Being women and the number of diagnosed diseases are factors that risk AA. On the other hand, the years of education are a protective factor for AA.

The results for the total sample highlights the importance of the absence of diseases and high educational background in achieving AA. However, educational background is significantly associated with AA in men, while in women remains marginally related.

Table 3.2: Predictive models of active ageing for the total sample and stratified by gender

Variables OR CI p-value

Total sample

Gender 1.82 (0.99-3.32) 0.051

Diseases 1.46 (1.21-1.78) <0.001

Education (years) 0.94 (0.91-0.98) <0.01 Marital status 0.96 (0.56-1.66) 0.896

BMI 1.03 (0.95-1.12) 0.493

Arm circumference 0.98 (0.91-1.06) 0.581 Waist circumference 1.00 (0.97-1.03) 0.785 Men

Diseases 1.69 (1.19-2.39) <0.005

Education (years) 0.94 (0.88-0.99) <0.01 Marital status 1.60 (0.50-5.11) 0.429

BMI 1.05 (0.88-1.24) 0.609

Arm circumference 0.95 (0.82-1.08) 0.417 Waist circumference 0.98 (0.93-1.03) 0.420 Women

Diseases 1.35 (1.06-1.72) <0.05

Education (years) 0.94 (0.89-1.00) 0.062 Marital status 0.88 (0.47-1.64) 0.680

BMI 1.02 (0.91-1.14) 0.802

Arm circumference 0.98 (0.89-1.08) 0.702 Waist circumference 1.02 (0.97-1.06) 0.494

32

(35)

3.1.2 Social disparities in health: the role of health be- haviours (Papers II and III)

In Papers II (4.2) and III (4.3) we describe SE inequalities in dynamic balance and frailty in Spain, respectively. Moreover, we analyse how different health behaviours like physical activity, obesity, alcohol and tobacco consumption, are associated with such health disparities. Both studies have a cross-sectional design and include data on Spanish individuals. Although Paper II (4.2) uses data from ELEA and Paper III (4.3) from SHARE, it is expected that some general conclusions may arise from both studies.

On the one hand, Paper II (4.2) investigates the association between SES and dynamic balance and whether lifestyle factors explained any possible associations. The sample is formed by 448 non-disabled individuals -age range 54-75-, enrolled in the ELEA project. We obtained an objective dichotomous (“good/poor”) measure of dynamic balance and used it as the dependent vari- able.

The main analysis in Table 3.3 shows that individuals with primary education had higher risk of poor dynamic balance, even after adjusting for age, gender, obesity, physical activity and income. In addition, obesity and sedentary/poor physical activity were related to poor dynamic balance.

Other results suggest a graded marginally significant association between household income and dynamic balance performance (not shown). However, the association is attenuated when adjusting for the covariates. Moreover, we also stratify the analysis by gender, as there are differences in balance, SES, health and lifestyles between men and women. However, estimates are similar for men and women in both education and income, suggesting that the associations between SES and dynamic balance is not modified by gender (data not shown).

On the other hand, Paper III (4.3) investigates educational differences in frailty phenotypes and whether obesity could explain any possible associations.

The study is based on 2,319 community-dwelling Spanish adults over 50 years old, participating in W1 of SHARE.

We defined frailty phenotypes based on the validated SHARE’s frailty criteria [52, 60]. Educational differences in frailty and its association with BMI -estimated by means of self-reports of height and weight- were evaluated using multinomial logistic regression analysis.

In line with previous results and compared to men, a larger proportion of

(36)

Table 3.3: ORs (95% CI) for poor dynamic balance by educational level, obesity, physical activity and income

Poor dynamic balance OR (95% IC) Educational level

Superior 1.00

High school 1.85 (0.81-4.22)

Primary 2.30 (1.16-4.56)

No formal education 1.62 (0.72-3.64) Obesity

No 1.00

Yes 1.49 (0.90-2.48)

Physical activity

Moderate / Vigorous 1.00

Light 1.45 (0.85-2.48)

Sedentary / Poor 2.89 (1.46-5.74) Notes: Adjusted for gender, age, education, obesity, physical activity and income.

women experienced frailty (22.3% vs. 13.3%; p<0.001 ) and pre-frailty (50.4%

vs. 46.5%; p<0.001), respectively. Figure 3.2 shows the distribution of frailty scores by BMI and stratified by educational background. Both underweight and obesity are associated with higher frailty scores, whereas there is an educational gradient in frailty scores (Figure 3.2).

Table 3.4 shows the main findings. After adjusting for all confounders, there is a significant educational gradient in frailty, where individuals with non-formal education show increased odds of a frailty phenotype than individuals with higher education. Moreover, obesity is significantly associated to frailty and the effect of the former is similar at all levels of education after testing for interaction effects (not shown).

3.1.3 Predictors of physical functional decline in Spanish adults (Paper IV)

In Paper IV (4.4) we evaluate the social, educational, health and behavioural predictors of physical FD between W1 and W2. This is a 2-year longitudinal study, based on 699 community-dwelling Spanish adults aged over 65 years

34

(37)

Figure 3.2: Distribution of frailty score (0-5) by BMI and educational level.

participating in SHARE.

We examine several baseline and longitudinal predictors of a combined measure of functional disability using logistic regressions. Table 3.5 shows the baseline factors associated with FD. There is an age dependent FD. There are gender differences in the association of baseline covariates and disability. Compared with women, older and poorly educated (low/primary) men are more likely to experience FD across waves. However, women reporting fair / poor health are more likely to experience FD than men.

Table 3.6 shows the decrease in general welfare in the covariates at follow-up, adjusting for confounders. FD in men is associated with an increased number of chronic diseases and depressive symptoms over a 2-year period, whereas among women it is associated with decreased numeracy score.

(38)

Table 3.4: ORs (95% CI) for pre-frailty and frailty phenotypes by educational level and BMI

Pre-Frail Frail

OR (95% CI) OR (95% CI) Education

ref. (High) 1.0 1.0

Medium 0.9 (0.6-1.2) 1.4 (0.6-3.3)

Low 1.1 (0.8-1.7) 1.9 (0.8-4.2)

Non-formal 1.2 (0.8-1.8) 2.3 (1.0-5.4) BMI

ref. (Normal) 1.0 1.0

Underweight 3.1 (0.3-28.2) 5.0 (0.4-58.6) Overweight 0.9 (0.7-1.1) 0.9 (0.6-1.3) Obesity 1.2 (0.9-1.7) 1.8 (1.2-2.7)

Adjusted for gender, age, education, chronic conditions, ADL, IADL, SRHS, smoke and alcohol consumption and BMI.

3.1.4 Growth under poor health: long-term health impli- cations in Europe (Paper V)

Based on data from SHARE, in Paper V (4.5) we analyse how poor health in childhood is associated with later life health within different European regions.

Moreover, we assess whether both the exposure to different SE situations over the life course and contemporary health behaviours mediated that association, over a 6.6 year window. To summarise the health status during childhood we use self-reports of general health in childhood, which has been described as a good predictor of adult health [64,73]. As indicators of adult health, we use self- reported health in adulthood (SRHS), the number of chronic diseases/conditions diagnosed by a physician and the handgrip strength (kg).

Descriptive statistics indicate that poor childhood health is relatively rare (10%) and women have worse health during childhood than men (p<0.05) (not shown). Western Europeans exhibit the highest proportion of poor health in childhood, and subjects with poor health in childhood show worse indicators of adult health, like poorer SRHS, higher chronic conditions and lower grip strength. Those with poor health in childhood are more likely to live as single and have fewer children. In addition, they are more likely to have the lowest educational achievement and the lowest incomes. Finally, those with poor health

36

(39)

Table 3.5: Baseline factors associated with functional decline. Multiple logistic regression analyses with OR and 95% CI.

Funct. decline OR (95% CI)

Variables Men Women Total

Sociodemographic Age (at baseline)

65-74 1.00 1.00 1.00

75-84 2.05 (1.10-3.84) 1.88 (1.14-3.08) 1.95 (1.32-2.87) 85+ 4.18 (1.45-12.02) 1.75 (0.76-4.04) 2.33 (1.22-4.48) Education

Medium/High 1.00 1.00 1.00

Low 2.48 (1.06-5.80) 1.19 (0.53-2.68) 1.22 (0.69-2.17) No formal educ. 1.25 (0.47-3.34) 0.63 (0.28-1.39) 1.47 (0.81-2.69) Self-reported health

SRHS

Exc./V. good 1.00 1.00 1.00

Good 0.76 (0.30-1.93) 3.87 (1.09-13.74) 1.60 (0.78-3.30) Fair 1.56 (0.62-3.93) 5.63 (1.63-19.49) 2.78 (1.37-5.62) Poor 1.69 (0.49-5.78) 5.37 (1.44-20.04) 2.65 (1.18-5.94) Symptoms^a 1.21 (1.01-1.46) 1.11 (0.99-1.24) 1.14 (1.03-1.25)

aThe number of medical symptoms present for at least the past 6 months at baseline;

adjusted for age, sex, years of education, occurrence of heart attack/stroke and depression.

in childhood are more likely to be physically inactive but non-smokers. In summary, several SE indicators over the life course and adult health behaviours are associated with childhood health, highlighting the relevance of adjusting for these variables in regression analysis.

Table 3.7 shows the regression estimates for the three adult outcomes. Only the final models for every outcome are shown. We found a direct association between self-reports of poor childhood health and worse adult health outcomes, even after controlling for all covariates. Compared to those with good health, subjects with poor health in childhood show significantly worse SRHS, higher chronic conditions and lower grip strength in adulthood (Table 3.7).

Furthermore, we found long-term gender differences in poor childhood health exposure. Growing under poor health is a stronger predictor of adult SRHS in women than in men.

Additionally, the exhibited gradient for Models 2 (not shown) might indicate

(40)

Table 3.6: Odds ratios (95% CI) for functional decline by decrease in general welfare at follow-up.

Funct. decline OR (95% CI)

Variables Men Women Total

Self-reported health

∆ SRHS

no change/improved 1.00 1.00 1.00

decreased 1.69 (0.90-3.14) 1.32 (0.81-2.16) 1.46 (1.00-2.19)

∆ Chronic diseases

no change/less 1.00 1.00 1.00

increased 2.25 (1.21-4.19) 1.33 (0.82-2.17) 1.63 (1.11-2.39)

∆ Symptoms

no change/less 1.00 1.00 1.00

increased 3.66 (1.94-6.88) 2.36 (1.45-3.86) 2.81 (1.91-4.12)

∆ BMI

worse 1.38 (0.59-3.24) 0.95 (0.49-1.83) 1.07 (0.63-1.79) Cognitive functioning/

mental health

∆ Orientation

decreased 1.59 (0.70-3.62) 1.22 (0.65-2.27) 1.31 (0.80-2.14)

∆ Numeracy score

decreased 1.71 (0.89-3.25) 1.88 (1.05-3.34) 1.81 (1.19-2.78)

∆ EURO-D

decreased 5.05 (2.42-10.54) 1.89 (1.08-3.29) 2.74 (1.78-4.22) Adjusted for age, sex, years of education, heart attack/stroke, depression and level of disability at baseline; SRHS: Self-reported health status, Orientation: Orientation test, Numeracy score: numeracy test, EURO-D: European Depression Scale.

∆: change across time (∆=W1-W2). If welfare status between W1 and W2 for all variables did not change/improved =0 (reference category); if decreases = 1.

Worse BMI was established when individuals change their BMI category in W1 to a worse BMI category in W2 normal to underweight or normal to overweight or normal to obese or overweight to obese.

that poor childhood health is a stronger predictor of adult health in Northern countries. Figure 3.3 illustrates the described pattern for the country region-

38

(41)

childhood health interaction for Models 2 (not shown).

Figure 3.3: Regression coefficients for poor childhood health on regional-specific adult health, represented by SRHS, chronic conditions and grip strength in W4. Estimates correspond to Model 2 and controls for demographic characteristics and early life conditions. *p<.05; **p<.01; ***p<.001

Finally, poor childhood health predicts declines in SRHS over 6.6 years, while controlling for covariates attenuated the associations for changes over time in chronic conditions and grip strength (data not shown).

3.2 General discussion

The aim of this study was to evaluate biological and social factors associated with different health outcomes among older adults, and to provide consistent data on integrating gender/sex considerations throughout the life course. This PhD study was based on two research projects, ELEA and SHARE, each con- sisting on a rather large sample of community-dwelling older people. I believe the data and the study designs were appropriate for investigating the research

(42)

questions, and that the original objectives were correctly addressed, by providing broad evidence on several determinants of adult health.

3.2.1 Discussion on the original papers

In Paper (4.1) we use a gender-based perspective for studying several factors associated to AA. The results of this paper might suggest distinctive associations for men and women. The differences observed in AA between men and women in relation to education are surely determined by a gender effect throughout the life course, mainly conditioned by access to education, home labours, etc. WHO’s recommendations on this highlights the importance of promoting education as a determinant factor to achieve AA [8]. However, greater knowledge is needed to less-known health conditions that may differ between men and women. The lower perceived and objective health status of women compared with men could be due to inequalities in the physical and psychological burden associated with the care of elders/disabled people or family members and must be considered differentially, since these activities are mainly conducted by women [74]. How- ever, there is a methodological issue that needs to be highlighted. There is no agreement on how to define AA, and also a controversy about the need to include elements that matter to older adults [75]. But independently on how to define it, there is also a debate about appropriate cut-off points in the measures used [12]. Some authors suggest that comparing the prevalences of active agers across studies may be of limited use given the wide variety of definitions and measurement approaches [12]. Moreover, when defining individuals into active and non-active agers, it is assumed that people cannot actively age with different chronic conditions [76]. For this reason, it may be appropriate to consider AA as a continuum, analysing how much someone ages actively would provide richer information than merely indicating whether someone is actively ageing or not. This approach, independently of the different dimensions and models used, would probably be a good initiative to board in future research.

In Papers II (4.2) and III (4.3) we addressed SE disparities in balance performance and frailty respectively, and whether health behaviours might explain such associations. To the well-known SE disparities in health in Spain, our studies add to the evidence on how health behaviours might contribute to those inequalities. Whereas some studies conclude that lifestyles make a relatively minor contribution to the social gradient in health [77], others have shown that differences in lifestyles can explain a relevant part of health inequalities [78,79].It

40

(43)

has been shown, e.g., that both physical activity and BMI reduced SE disparities in functioning to some extent [80, 81] differently affecting men and women [82].

However, our study (Paper II (4.2)) does not show significant reduction in SE disparities in dynamic balance when controlling for obesity or physical activity, suggesting an independent effect of both SES and behavioural factors on balance performance. Similarly, in Paper III (4.3) we found that the effect of obesity does not vary within levels of education, which makes the contribution of the former to the odds of frailty phenotype somehow independent of the educational background, although there is a somehow mediation of obesity. The relevance of these findings relies on contributing to assess the importance of health behaviours in the magnitude of health inequalities. It has been suggested that risk factors are unequally distributed among the social classes and serve as potential pathways through which SES may influence health in older adults [83].

Our findings suggest that health behaviours like physical inactivity and obesity are not enough to explain the SE disparities in the studied outcomes, although no further considerations can be done due to the cross-sectional designs of the studies. The large SE differences in health in Spanish older adults make this type of studies very important in order to tackle health inequalities among older men and women, and longitudinal approaches are essential to accomplish that in future work.

In Paper IV (4.4) we studied the predictors of physical functional disability in older Spanish adults. We found that longitudinal changes in predictors are strongly associated with longitudinal changes in function between baseline and a 2-year follow up, most clearly among men. Although men reported better health than women, changes in SRHS were more strongly associated with decline in men. One possible interpretation is that health and symptoms are already so much worse for women than for men at baseline, that longitudinal changes for women are smaller than for men -due to ceiling effects. As a result, at any given age, change may appear to be a stronger predictor for the relatively healthier men, while the already poor levels of health at baseline may turn out to be stronger predictors for women. Our findings suggest that in addition to age, education and SRHS, the onset of symptoms, the onset of chronic diseases and depressive symptoms, and reduced cognitive functioning (numeracy score) are clinical predictors potentially useful in the prevention of disability in older Spanish adults. The prevention/delay of the onset of disability in older adults could have a positive impact on their quality of life [84]. Effective strategies are needed for the prevention of FD and our study provides tools for its clinical prediction, which may help reducing the incidence of disabilities and the period

(44)

of dependence near the end of life and curb increasing trends in disability in the older Spanish population.

Finally, our study on the life course trajectories of health (Paper V (4.5)) showed interesting results. Overall, our results indicate that the impact of childhood health in adult health is direct, whereas the impact of childhood SES is more indirect, operating through own SES in adulthood. Our results are not inconsistent with what much of the previous literature stands, adding to the evidence on the importance of early life health in determining later life health [63–65, 85]. Based on this, our results support the notion that in order to improve old-age health, efficient interventions could be guided to first improve child health. Moreover, we found long-term gender differences in poor childhood health exposure. Growing under poor health is a stronger predictor of adult SRHS in women than in men, and our results may suggest that part of the excess in worse subjective health among women relative to men might have an early origin due to the traditional gender roles established early in life. This adds to the evidence of previous reports on gender health inequalities over the life course [16]. Finally, our findings suggest that the early exposure to poor health states may have more negative impact in Northern compared to Western and Southern European countries, as shown in Figure 3.3. To our knowledge, these findings provide the first regional-specific evidence of historical differences in the childhood-adult health association. Unfortunately, we cannot offer a straightforward substantive explanation for this result. What this finding may suggest, however, is that growing under poor health in Southern and Western countries would be more random and therefore, independent of SES, while in Northern countries this would represent a highly persistent state, involving being sickly throughout the life course, but this is largely speculative. Comparative studies have found that socioeconomic inequalities in mortality and morbidity are not smaller in countries with relatively universal and generous welfare policies (like Nordic countries) than they are in other countries (e.g. Southern European countries with their more family-based welfare arrangements) [86]. It has sometimes been argued that advanced welfare states may raise unrealistic expectations of a better life among people with a lower SE position -and consequently, poorer health-, and therefore induce higher levels of frustration and stress [86, 87]. We cannot exclude for interpretations, however, a possible bias due to cross-cultural differences in self-reported health. The image of “normal”

health might differ among different societies [88] and this cross-cultural difference makes the consistency of the findings from international studies somehow less straightforward.

42

(45)

3.2.2 Implications and future directions

I believe that the results presented in the above mentioned studies may be useful for policy-makers when developing interventions focused on promoting health and preventing disability in older adults. Our five studies contribute to this in different ways. Although the cross-sectional designs of Papers I-III (4.1 4.2 4.3) do not allow to suggest causality and consequently, no further conclusions more than simple associations can be derived from them, findings from Papers IV (4.4) and V (4.5) provide more solid evidence that do enable to suggest interventions based on longitudinal evidences. Particularly, our results add to the evidence on the need to continue applying life course and gender-based approaches with longitudinal designs whenever possible in studies of health disparities among older adults. Our results and the growing evidence on the importance of the life course determinants of adult health, together with the historical gender and SE differences in access to education, income and healthy lifestyles, support these approaches.

It is worth mentioning that the results and possible implications presented here should be carefully considered. Thus, e.g., results derived from longitudinal analysis referring causality of the potential loss of functional capacity and the effect of childhood into adult health must be interpreted with caution. Overall and despite the limitations described in every paper, this study makes a valuable contribution to the knowledge on the SE determinants of adult health, by pre- senting gender-based evidences of disparities and health trajectories throughout the life course.

Future directions in life course research include the use of biomarkers as indicators of adult health. Physical and biological measurements as objective health data were so far mostly taken in smaller, non-representative clinical studies. But in the last couple of years more and more large-scale surveys added biomarkers to their programs, since there is a promising scientific value to it.

The use of biomarkers will enable researchers to i ) validate respondents’ self- reports, ii ) identify causal relationships and specific physiological pathways and help understanding the complex relationships between social status and health and iii ) give pre-disease information: physiological processes are often below the individual’s threshold of perception, but may be nevertheless predictive for ongoing or future diseases. [89]. At this moment, SHARE has released a pilot test in the German sub-sample and it is expected that in further waves it will cover the entire sample.

The understanding of gender inequalities in health is largely disconnected from the life course processes that precede them and the social contexts in which

(46)

they unfold [90]. We contribute to address this, by describing long-term gender differences in poor childhood health exposure that suggest that part of the excess in worse subjective health among women relative to men might have an early origin due to the traditional gender roles established early in life. Future work would benefit from greater sensitivity to the role of gender and its intersection with life course experiences [91].

44

(47)

Table 3.7: Estimates for adult self-reported health, chronic conditions and grip strength. Only final models are represented.

Covariates SRHS Diseases Grip strength

Health status when 10y(ref.=Good)

Poor 0.474*** 0.229** -1.542^a

Sex (ref.=Men) Women 0.027 0.080*** -15.811***

Health status when 10y*Sex (ref.=Good/Men)

Poor*Women 0.164* 0.067 -0.573

Age at first interview 0.017*** 0.023*** -0.469***

Countries (ref.=Northern)

Western 0.431*** 0.013 -1.901***

Southern 0.644*** 0.147a -5.617***

Health status when 10y*Countries (ref.=Good/Northern)

Poor*Western -0.278* -0.054 0.732

Poor*Southern -0.318* -0.204* 1.285

Childhood SES (ref.=Other)Low 0.121^a 0.059 0.307

Area of residence*Countries (ref.=Big city-subs./Northern)

Rural area/Village*Southern -0.028 0.134 2.126**

Living situation

(ref.=Living with spouse/partner)

Living as single 0.044 0.042^a -0.519*

Number of children 0.007 0.011 0.127*

Education (ref.=5^th quintile)

4^th quintile -0.011 -0.047 -0.379

3^rd quintile 0.063 -0.017 -0.116

2^ndquintile 0.156** 0.033 -0.829*

1^stquintile 0.206*** 0.014 -0.723*

Income (ref.=5^thquintile)

4^th quintile 0.093** 0.083* 0.069

3^rd quintile 0.149*** 0.053 -0.394

2^ndquintile 0.207*** 0.145*** -0.620*

1^stquintile 0.250*** 0.112** -0.745*

Obesity(BMI>30)(ref.=Other)Yes 0.275*** 0.334*** 0.050 Physical inact.(ref.=Other)Yes 0.447*** 0.141*** -2.473***

Smoke(ref.=Other)Yes 0.087*** 0.059** 0.047

Notes: Linear model for SRHS and grip strength; negative binomial model for chronic conditions.^ap<.10; *p<.05; **p<.01; ***p<.001.

(48)

Biosocial determinants of healthy ageing in Spain

PhD Thesis

Biosocial determinants of healthy ageing in Spain

Author:

Santiago Rodr´ıguez L´ opez

Supervisor:

Pilar Montero L´ opez

Departamento de Biolog´ıa Facultad de Ciencias

Universidad Aut´ onoma de Madrid ESPA ˜ NA

2014

Abstract

Abstract (Spanish)

Acknowledgements

List of original publications

Abbreviations

Contents

Introduction

1.1 Population ageing and the health paradox in Spain

1.2 Studying healthy ageing: Active Ageing and its bio/psycho/social frame

1.3 Gender/sex integrated aspects in health and welfare research

1.4 The life course perspective when studying health in adulthood

1.5 New contributions to the study of the de- terminants of health and well-being in older adults

Objectives

Material and Methods

2.1 Study design and participants

2.1.1 Active Ageing Longitudinal Study: the ELEA project

2.1.2 The Survey of Health, Ageing and Retirement in Europe (SHARE)

2.2 Outcomes

2.3 Statistical analysis

Results and Discussion

3.1 Summary of the results

3.1.1 Gender differences in active ageing in Spain (Paper I)

3.1.2 Social disparities in health: the role of health be- haviours (Papers II and III)

3.1.3 Predictors of physical functional decline in Spanish adults (Paper IV)

3.1.4 Growth under poor health: long-term health impli- cations in Europe (Paper V)

3.2 General discussion

3.2.1 Discussion on the original papers

3.2.2 Implications and future directions